Doing Udacity ML course. After df_final.join(df_temp, how="left") get NaN, but in the course venv everything works great. Where might be the problem?
P.S.: I also tried df_temp.index = pd.to_datetime(df_temp.index, utc=True) for each, seems no effect.
Here we load data.
import yfinance as yf
tickets = ["AAPL", "AMD", "GOOG", "GLD"]
def download_tickets(tickets):
for ticket in tickets:
df = yf.Ticker(ticket)
df = df.history(period="max")
df.to_csv(symbol_to_path(ticket))
Here we create path to csv from symbol.
def symbol_to_path(symbol, base_dir="data"):
if not os.path.exists(base_dir):
os.mkdir(base_dir)
return os.path.join(base_dir, "{}.csv".format(str(symbol)))
Here we join data.
# Create empty df with specified dates.
start_date = "2022-01-01"
end_date = "2023-01-01"
dates = pd.date_range(start_date, end_date)
df_final = pd.DataFrame(index=dates)
df_final.index = pd.to_datetime(df_final.index, utc=True)
# Combine all with df_final
for ticket in tickets:
file_path = symbol_to_path(symbol)
df_temp = pd.read_csv(file_path, parse_dates=True, index_col="Date",
usecols=["Date", "Close"], na_values=["nan"])
df_temp = df_temp.rename(columns={"Close": symbol})
df_final = df_final.join(df_temp, how="left")
print(df_temp.head())
print(df_final.head())
return df_final
Output:
As you see, float converts to NaN for left join
For right join we get data, but not for the range 2022-01-01/2023-01-01
Inner join
Outer join
Thank you.
UPD: Data after 2021
The problem is in time zones. Tickets data is in -05:00 (I assume new york), while you generate df_final at UTC +00:00, when you join, pandas cannot find intersection in indices.
Simplest solution for me was to change df_final timezone (tz), ie generate with correct tz
# Create empty df with specified dates.
start_date = "2022-01-01"
end_date = "2023-01-01"
dates = pd.date_range(start_date, end_date, tz='-05:00') # change here
df_final = pd.DataFrame(index=dates)
# df_final.index = pd.to_datetime(df_final.index, utc=True) # NOT needed anymore
Related
The title almost says it already. I have a pyspark.sql.dataframe.Dataframe with a "ID", "TIMESTAMP", "CONSUMPTION" and "TEMPERATURE" column. I need the "TIMESTAMP" column to be resampled to daily intervals (from 15min intervals) and the "CONSUMPTION" and "TEMPERATURE" column aggregated by summation. However, this needs to be performed for each unique id in the "ID" column. How do I do this?
Efficiency/speed is of importance to me. I have a huge dataframe to start with, which is why I would like to avoid .toPandas() and for loops.
Any help would be greatly appreciated!
The following code will build a spark_df to play around with. The input_spark_df represents the input spark dataframe, the disred output is like desired_outcome_spark_df.
import pandas as pd
import numpy as np
from pyspark.sql import SparkSession
df_list = []
for unique_id in ['012', '345', '678']:
date_range = pd.date_range(pd.Timestamp('2022-12-28 00:00'), pd.Timestamp('2022-12-30 23:00'),freq = 'H')
df = pd.DataFrame()
df['TIMESTAMP'] = date_range
df['ID'] = unique_id
df['TEMPERATURE'] = np.random.randint(1, 10, df.shape[0])
df['CONSUMPTION'] = np.random.randint(1, 10, df.shape[0])
df = df[['ID', 'TIMESTAMP', 'TEMPERATURE', 'CONSUMPTION']]
df_list.append(df)
pandas_df = pd.concat(df_list)
spark = SparkSession.builder.getOrCreate()
input_spark_df = spark.createDataFrame(pandas_df)
desired_outcome_spark_df = spark.createDataFrame(pandas_df.set_index('TIMESTAMP').groupby('ID').resample('1d').sum().reset_index())
To condense the question thus: how do I go from input_spark_df to desired_outcome_spark_df as efficient as possible?
I found the answer to my own question. I first change the timestamp to "date only" using pyspark.sql.functions.to_date. Then I groupby both "ID" and "TIMESTAMP" and perfrom the aggregation.
from pyspark.sql.functions import to_date, sum, avg
# Group the DataFrame by the "ID" column
spark_df = input_spark_df.withColumn('TIMESTAMP', to_date(col('TIMESTAMP')))
desired_outcome = (input_spark_df
.withColumn('TIMESTAMP', to_date(col('TIMESTAMP')))
.groupBy("ID", 'TIMESTAMP')
.agg(
sum(col("CONSUMPTION")).alias("CUMULATIVE_DAILY_POWER_CONSUMPTION"),
avg(col('TEMPERATURE')).alias("AVERAGE_DAILY_TEMPERATURE")
))
grouped_df.display()
I have a list of 16 dataframes that contain stats for each player in the NBA during the respective season. My end goal is to run unsupervised learning algorithms on the data frames. For example, I want to see if I can determine a player's position by their stats or if I can determine their total points during the season based on their stats.
What I would like to do is modify the list(df_list), unless there's a better solution, of these dataframes instead modifying each dataframe to:
Change the datatype of the MP(minutes played column from str to int.
Modify the dataframe where there are only players with 1000 or more MP and there are no duplicate players(Rk)
(for instance in a season, a player(Rk) can play for three teams in a season and have 200MP, 300MP, and 400MP mins with each team. He'll have a column for each team and a column called TOT which will render his MP as 900(200+300+400) for a total of four rows in the dataframe. I only need the TOT row
Use simple algebra with various and individual columns columns, for example: being able to total the MP column and the PTS column and then diving the sum of the PTS column by the MP column.
Or dividing the total of the PTS column by the len of the PTS column.
What I've done so far is this:
Import my libraries and create 16 dataframes using pd.read_html(url).
The first dataframes created using two lines of code:
url = "https://www.basketball-reference.com/leagues/NBA_1997_totals.html"
ninetysix = pd.read_html(url)[0]
HOWEVER, the next four data frames had to be created using a few additional line of code(I received an error code that said "html5lib not found, please install it" so I downloaded both html5lib and requests). I say that to say...this distinction in creating the DF may have to considered in a solution.
The code I used:
import requests
import uuid
url = 'https://www.basketball-reference.com/leagues/NBA_1998_totals.html'
cookies = {'euConsentId': str(uuid.uuid4())}
html = requests.get(url, cookies=cookies).content
ninetyseven = pd.read_html(html)[0]
These four data frames look like this:
I tried this but it didn't do anything:
df_list = [
eightyfour, eightyfive, eightysix, eightyseven,
eightyeight, eightynine, ninety, ninetyone,
ninetytwo, ninetyfour, ninetyfive,
ninetysix, ninetyseven, ninetyeight, owe_one, owe_two
]
for df in df_list:
df = df.loc[df['Tm'] == 'TOT']
df = df.copy()
df['MP'] = df['MP'].astype(int)
df['Rk'] = df['Rk'].astype(int)
df = list(df[df['MP'] >= 1000]['Rk'])
df = df[df['Rk'].isin(df)]
owe_two
============================UPDATE===================================
This code will solves a portion of problem # 2
url = 'https://www.basketball-reference.com/leagues/NBA_1997_totals.html'
dd = pd.read_html(url)[0]
dd = dd[dd['Rk'].ne('Rk')]
dd['MP'] = dd['MP'].astype(int)
players_1000_rk_list = list(dd[dd['MP'] >= 1000]['Rk'])
players_dd = dd[dd['Rk'].isin(players_1000_rk_list)]
But it doesn't remove the duplicates.
==================== UPDATE 10/11/22 ================================
Let's say I take rows with values "TOT" in the "Tm" and create a new DF with them, and these rows from the original data frame...
could I then compare the new DF with the original data frame and remove the names from the original data IF they match the names from the new data frame?
the problem is that the df you are working on in the loop is not the same df that is in the df_list. you could solve this by saving the new df back to the list, overwriting the old df
for i,df in enumerate(df_list):
df = df.loc[df['Tm'] == 'TOT']
df = df.copy()
df['MP'] = df['MP'].astype(int)
df['Rk'] = df['Rk'].astype(int)
df = list(df[df['MP'] >= 1000]['Rk'])
df = df[df['Rk'].isin(df)]
df_list[i] = df
the2 lines are probably wrong as well
df = list(df[df['MP'] >= 1000]['Rk'])
df = df[df['Rk'].isin(df)]
perhaps you want this
for i,df in enumerate(df_list):
df = df.loc[df['Tm'] == 'TOT']
df = df.copy()
df['MP'] = df['MP'].astype(int)
df['Rk'] = df['Rk'].astype(int)
#df = list(df[df['MP'] >= 1000]['Rk'])
#df = df[df['Rk'].isin(df)]
# just the rows where MP > 1000
df_list[i] = df[df['MP'] >= 1000]
I need to get top n rows by some value per week (and I have hourly data).
data:
import numpy as np
import pandas as pd
dates = pd.date_range(start='1/1/2020', end='11/1/2020', freq="1H")
values = np.random.randint(20, 100500, len(dates))
some_other_column = np.random.randint(0, 10000000, len(dates))
df = pd.DataFrame({"date": dates, "value": values, "another_column": some_other_column})
My attempt:
resampled = df.set_index("date").resample("W")["value"].nlargest(5).to_frame()
It does give top 5 rows but all other columns except for date and value are missing - and I want to keep them all (in my dataset I have lots of columns but here another_column just to show that it's missing).
The solution I came up with:
resampled.index.names = ["week", "date"]
result = pd.merge(
resampled.reset_index(),
df,
how="left",
on=["date", "value"]
)
But it all feels wrong, I know there should be much simpler solution. Any help?
The output I was looking for. Thanks #wwnde.
df["week"] = df["date"].dt.isocalendar().week
df.loc[df.groupby("week")["value"].nlargest(5).index.get_level_values(1), :]
Groupby, and mask any nlargest
df.set_index('date', inplace=True)
df[df.groupby(df.index.week)['value'].transform(lambda x:x.nlargest(5).any())]
My data is as below. I want to sort by the timestamp and use the latest sample of each userid as the testing data. How should I do the train and test split? What I have tried is using pandas to sort_values timestamp and then groupby 'userid'. But I only get a groupby object. What is the correct way to do that? Is pyspark a better tool?
After I get the dataframe of the testing data, how should split data? Obviously I cannot use sklearn's train_test_split.
You could do the following:
# Sort the data by time stamp
df = df.sort_values('timestamp')
# Group by userid and get the last entry from each group
test_df = df.groupby(by='userid', as_index=False).nth(-1)
# The rest of the values
train_df = df.drop(test_df.index)
You can do the following:
import pyspark.sql.functions as F
max_df = df.groupby("userid").agg(F.max("timestamp"))
# join it back to the original DF
df = df.join(max_df, on="userid")
train_df = df.filter(df["timestamp"] != df["max(timestamp)"])
test_df = df.filter(df["timestamp"] == df["max(timestamp)"])
I asked (and answered) a question here Pandas ffill resampled data grouped by column where I wanted to know how to ffill a date range for each unique entry for a column (my assets column).
My solution requires that the asset "id" is a column. However, the data makes more sense to me as a multiindex. Furthermore I would like more fields in the multiindex. Is the only way of filling forward to drop the non-date fields from the multiiindex before ffilling?
A modified version of my example (to work on a df with multiindex) here:
from datetime import datetime, timedelta
import pytz
some_time = datetime(2018,4,2,20,20,42)
start_date = datetime(some_time.year,some_time.month,some_time.day).astimezone(pytz.timezone('Europe/London'))
end_date = start_date + timedelta(days=1)
start_date = start_date + timedelta(hours=some_time.hour,minutes=(0 if some_time.minute < 30 else 30 ))
df = pd.DataFrame(['A','B'],columns=['asset_id'])
df2=df.copy()
df['datetime'] = start_date
df2['datetime'] = end_date
df['some_property']=0
df.loc[df['asset_id']=='B','some_property']=2
df = df.append(df2).set_index(['asset_id','datetime'])
With what is arguably my crazy solution here:
df = df.reset_index()
df = df.set_index('datetime').groupby('asset_id').resample('30T').ffill().drop('asset_id',axis=1)
df = df.reset_index().set_index(['asset_id','datetime'])
Can I avoid all that re-indexing?